TenantAtlas/specs/339-provider-connection-scope-hardening/plan.md

# Implementation Plan: Spec 339 - Provider Connection Scope Hardening

- Branch: `339-provider-connection-scope-hardening`
- Date: 2026-05-31
- Spec: `specs/339-provider-connection-scope-hardening/spec.md`
- Input: Spec 339 + repo inspection of Provider Connection policy/resource authority seams.

## Summary

Harden Provider Connections (credential-adjacent) so that authorization and scope authority come only from:

- explicit workspace context (`WorkspaceContext`, session),
- explicit `environment_id` (filter/create), or
- record ownership (workspace + managed environment) for view/edit/actions.

Remove hidden fallbacks (remembered environment and Filament tenant) as authority sources and guard the contract with targeted tests.

## Technical Context

- Language/Version: PHP 8.4.15, Laravel 12.52.x.
- Primary Dependencies: Filament 5.x, Livewire 4.x, Pest 4.x.
- Storage: PostgreSQL; no schema change expected.
- Testing: Pest Feature tests (ProviderConnections family).
- Validation Lanes: fast-feedback (Feature). Browser smoke only if visible UX changes are introduced.
- Target Platform: Sail locally; Dokploy/container posture unchanged.
- Project Type: Laravel monolith under `apps/platform`.
- Constraints: no new persisted truth, migrations, packages, or broad navigation redesign.

## UI / Surface Guardrail Plan

- **Guardrail scope**: existing operator-facing resource + credential-adjacent action gating (authority hardening).
- **Affected surfaces**:
  - Provider Connections list/view/create/edit routes (`/admin/provider-connections*`).
  - Create CTA availability and create authorization (must require explicit `environment_id`).
  - Policy semantics (404 vs 403) and record-derived authority for actions.
- **Native vs custom**: native Filament + existing capability/policy seams; no custom UI framework.
- **Shared-family relevance**: workspace hub filter contract, authorization semantics, action-surface discipline for credential-adjacent actions.
- **State layers in scope**: URL query (`environment_id` only), session workspace context, record ownership.
- **Handling modes**: review-mandatory (security semantics).
- **Required tests / smoke**:
  - Feature tests that prove forbidden fallbacks cannot authorize create/view/actions.
  - Optional reuse of existing browser smoke only if UI copy/CTA visibility changes are required.
- **UI/Productization coverage**: no new surface; do not update UI audit coverage files unless navigation routes change.

## Shared Pattern & System Fit

- **Cross-cutting feature marker**: yes.
- **Systems touched (expected)**:
  - `apps/platform/app/Policies/ProviderConnectionPolicy.php`
  - `apps/platform/app/Filament/Resources/ProviderConnectionResource.php`
  - `apps/platform/app/Support/Workspaces/WorkspaceContext.php`
  - `apps/platform/app/Support/Navigation/WorkspaceHubEnvironmentFilter.php`
  - `apps/platform/tests/Feature/ProviderConnections/*`
- **Shared abstractions reused**: `WorkspaceContext`, deny-as-not-found semantics, capability resolver and gates, existing ProviderConnections test family.
- **New abstraction introduced?**: none.
- **Bounded deviation / spread control**: do not introduce a new “scope resolver framework”; keep changes local to the ProviderConnection policy/resource seams.

## OperationRun UX Impact

No new `OperationRun` types or start UX changes.

Provider operations reachable from Provider Connections must remain scoped to the record’s owning environment/workspace, but this spec’s changes should be limited to authority and authorization guardrails.

## Implementation Approach

### Phase 1 — Repo truth + failing tests first

- Inventory the current forbidden fallbacks:
  - `ProviderConnectionPolicy::currentWorkspace()` fallback to `Filament::getTenant()`.
  - `ProviderConnectionPolicy::resolveCreateTenant()` fallback to `WorkspaceContext::lastEnvironmentId()` and `Filament::getTenant()`.
  - Provider Connection resource query scoping fallback paths when workspace context is missing.
- Add failing tests proving:
  - create is impossible without explicit `environment_id` (even if remembered environment exists),
  - Filament tenant cannot substitute for workspace context,
  - legacy query aliases do not widen access or grant authority.

### Phase 2 — Policy authority hardening

- Remove forbidden fallbacks in `ProviderConnectionPolicy`:
  - workspace authority must come from `WorkspaceContext` only,
  - create tenant must come from explicit `environment_id` only.
- Keep 404 vs 403 semantics consistent with repo rules.

### Phase 3 — Resource scoping hardening

- Ensure `ProviderConnectionResource` list/query scoping:
  - requires workspace context for membership scope (no tenant-context fallback),
  - uses `environment_id` only as an explicit filter, validated against workspace/access scope.
- Ensure record pages/actions always use record-derived tenant/workspace (no hidden context authority).

### Phase 4 — Validation and regression posture

Run narrow tests first:

- `cd apps/platform && ./vendor/bin/sail artisan test --compact tests/Feature/ProviderConnections --filter=ScopeHardening`
- `cd apps/platform && ./vendor/bin/sail pint --dirty --format agent`

Run optional browser smoke only if implementation changes visible CTA/copy:

- `cd apps/platform && ./vendor/bin/sail artisan test --compact tests/Browser/Spec281ProviderConnectionScopeSmokeTest.php`

## Deployment / Ops Impact

- Migrations: none.
- Env vars: none.
- Queues/scheduler: none (no new job orchestration beyond existing provider operations).
- Filament assets: unchanged.

> **Fill when the feature touches shared provider/platform seams, identity scope, governed-subject taxonomy, compare strategy selection, provider connection descriptors, or operator vocabulary that may leak provider-specific semantics into platform-core truth. Docs-only or template-only work may use concise `N/A`.**

- **Shared provider/platform boundary touched?**: [yes / no / N/A]
- **Provider-owned seams**: [List or `N/A`]
- **Platform-core seams**: [List or `N/A`]
- **Neutral platform terms / contracts preserved**: [List or `N/A`]
- **Retained provider-specific semantics and why**: [none / short explanation]
- **Bounded extraction or follow-up path**: [none / document-in-feature / follow-up-spec / N/A]

## Constitution Check

*GATE: Must pass before Phase 0 research. Re-check after Phase 1 design.*

- Inventory-first: clarify what is “last observed” vs snapshots/backups
- Read/write separation: any writes require preview + confirmation + audit + tests
- Graph contract path: Graph calls only via `GraphClientInterface` + `config/graph_contracts.php`
- Deterministic capabilities: capability derivation is testable (snapshot/golden tests)
- RBAC-UX: two planes (/admin vs /system) remain separated; cross-plane is 404; tenant-context routes (/admin/t/{tenant}/...) are tenant-scoped; canonical workspace-context routes under /admin remain tenant-safe; non-member tenant/workspace access is 404; member-but-missing-capability is 403; authorization checks use Gates/Policies + capability registries (no raw strings, no role-string checks)
- Workspace isolation: non-member workspace access is 404; tenant-plane routes require an established workspace context; workspace context switching is separate from Filament Tenancy
- RBAC-UX: destructive-like actions require `->requiresConfirmation()` and clear warning text
- RBAC-UX: global search is tenant-scoped; non-members get no hints; inaccessible results are treated as not found (404 semantics)
- Tenant isolation: all reads/writes tenant-scoped; cross-tenant views are explicit and access-checked
- Run observability: long-running/remote/queued work creates/reuses `OperationRun`; start surfaces enqueue-only; Monitoring is DB-only; DB-only <2s actions may skip runs but security-relevant ones still audit-log; auth handshake exception OPS-EX-AUTH-001 allows synchronous outbound HTTP on `/auth/*` without `OperationRun`
- OperationRun start UX: any feature that creates, queues, deduplicates, resumes, blocks, completes, or links `OperationRun` reuses the central OperationRun Start UX Contract; no local composition of queued toast/link/event/start-state messaging; `OperationRun UX Impact` is present in the active spec or plan
- Ops-UX 3-surface feedback: if `OperationRun` is used, default feedback is toast intent-only + progress surfaces + exactly-once terminal `OperationRunCompleted` (initiator-only); queued DB notifications remain explicit opt-in through the shared start UX contract; running DB notifications stay disallowed
- Ops-UX lifecycle: `OperationRun.status` / `OperationRun.outcome` transitions are service-owned (only via `OperationRunService`); context-only updates allowed outside
- Ops-UX summary counts: `summary_counts` keys come from `OperationSummaryKeys::all()` and values are flat numeric-only
- Ops-UX guards: CI has regression guards that fail with actionable output (file + snippet) when these patterns regress
- Ops-UX system runs: initiator-null runs emit no terminal DB notification; audit remains via Monitoring; tenant-wide alerting goes through Alerts (not OperationRun notifications)
- Automation: queued/scheduled ops use locks + idempotency; handle 429/503 with backoff+jitter
- Data minimization: Inventory stores metadata + whitelisted meta; logs contain no secrets/tokens
- Test governance (TEST-GOV-001): actual test-purpose classification, affected lanes, fixture/helper/factory/seed/context cost risks, heavy-family visibility, review-stop points, reviewer handoff, and any budget/baseline/trend follow-up are explicit; the narrowest proving lane mix is planned and any structural cost change has an escalation path
- Proportionality (PROP-001): any new structure, layer, persisted truth, or semantic machinery is justified by current release truth, current operator workflow, and why a narrower solution is insufficient
- No premature abstraction (ABSTR-001): no new factories, registries, resolvers, strategy systems, interfaces, type registries, or orchestration pipelines before at least 2 real concrete cases exist, unless security, tenant isolation, auditability, compliance evidence, or queue correctness require it now
- Persisted truth (PERSIST-001): new tables/entities/artifacts represent independent product truth or lifecycle; convenience projections and UI helpers stay derived
- Behavioral state (STATE-001): new states/statuses/reason codes change behavior, routing, permissions, lifecycle, audit, retention, or retry handling; presentation-only distinctions stay derived
- UI semantics (UI-SEM-001): avoid turning badges, explanation text, trust/confidence labels, or detail summaries into mandatory interpretation frameworks; prefer direct domain-to-UI mapping
- Shared pattern first (XCUT-001): cross-cutting interaction classes reuse existing shared contracts/presenters/builders/renderers first; any deviation is explicit, bounded, and justified against current-release truth
- Provider boundary (PROV-001): shared provider/platform seams are classified as provider-owned vs platform-core; provider-specific semantics stay out of platform-core contracts, taxonomy, identifiers, compare semantics, and operator vocabulary unless explicitly justified; bounded extraction beats speculative multi-provider frameworks
- V1 explicitness / few layers (V1-EXP-001, LAYER-001): prefer direct implementation, local mappings, and small helpers; any new layer replaces an old one or proves the old one cannot serve
- Spec discipline / bloat check (SPEC-DISC-001, BLOAT-001): related semantic changes are grouped coherently, and any new enum, DTO/presenter, persisted entity, interface/registry/resolver, or taxonomy includes a proportionality review covering operator problem, insufficiency, narrowness, ownership cost, rejected alternative, and whether it is current-release truth
- Badge semantics (BADGE-001): status-like badges use `BadgeCatalog` / `BadgeRenderer`; no ad-hoc mappings; new values include tests
- Filament-native UI (UI-FIL-001): admin/operator surfaces use native Filament components or shared primitives first; no ad-hoc status UI, local semantic color/border decisions, or hand-built replacements when native/shared semantics exist; any exception is explicitly justified
- Filament-native UI (UI-FIL-001): custom Filament UI follows `docs/ui/tenantpilot-enterprise-ui-standards.md`; no ad-hoc custom styling for cards, buttons, hovers, badges, icons, progress bars, empty states, or interactive rows
- Filament-native UI (UI-FIL-001): if local Blade/Tailwind cards are
  still necessary, they preserve dark mode correctness, spacing
  consistency, badge semantics, action hierarchy, progressive
  disclosure, accessibility, and Filament visual language
- Filament-native UI (UI-FIL-001): custom Blade/Widget/Page surfaces
  keep Filament-native interaction semantics, preserve one dominant
  primary action per focused area, express state through
  BADGE-001-aligned badges/labels/supporting text instead of arbitrary
  button colors, and do not create independent button/card/spacing
  systems
- Filament-native UI (UI-FIL-001): hover, pointer, focus, shadow, or
  similar interactive affordance appears only when a repo-real
  route/action and permitted capability exist; non-interactive rows
  stay visibly static
- UI/UX surface taxonomy (UI-CONST-001 / UI-SURF-001): every changed operator-facing surface is classified as exactly one allowed surface type; ad-hoc interaction models are forbidden
- Decision-first operating model (DECIDE-001): each changed
  operator-facing surface is classified as Primary Decision,
  Secondary Context, or Tertiary Evidence / Diagnostics; primary
  surfaces justify the human-in-the-loop moment, default-visible info
  is limited to first-decision needs, deep proof is progressive
  disclosed, one governance case stays decidable in one context where
  practical, navigation follows workflows not storage structures, and
  automation / alerts reduce attention load instead of adding noise
- Audience-aware disclosure (DECIDE-AUD-001 / OPSURF-001): detail or
  status surfaces separate customer-readable decision content,
  operator diagnostics, and support/raw evidence; customer-readable
  default paths hide raw JSON, copied context, fingerprints, internal
  reason ownership, platform reason families, and debug semantics;
  one dominant next action is explicit; duplicate visible truth is
  removed
- UI/UX inspect model (UI-HARD-001): each list surface has exactly one primary inspect/open model; redundant View beside row click or identifier click is forbidden; edit-as-inspect is limited to Config-lite resources
- UI/UX action hierarchy (UI-HARD-001 / UI-EX-001): standard CRUD and Registry rows expose at most one inline safe shortcut; destructive actions are grouped or in the detail header; queue exceptions are catalogued, justified, and tested
- UI/UX scope, truth, and naming (UI-HARD-001 / UI-NAMING-001 / OPSURF-001): scope signals are truthful, canonical nouns stay stable across shells, critical operational truth is default-visible, and standard lists remain scanable
- UI/UX placeholder ban (UI-HARD-001): empty `ActionGroup` / `BulkActionGroup` placeholders and declaration-only UI conformance are forbidden
- UI naming (UI-NAMING-001): operator-facing labels use `Verb + Object`; scope (`Workspace`, `Tenant`) is never the primary action label; source/domain is secondary unless disambiguation is required; runs/toasts/audit prose use the same domain vocabulary; implementation-first terms do not appear in primary operator UI
- Operator surfaces (OPSURF-001): `/admin` defaults are operator-first; default-visible content avoids raw implementation detail; diagnostics are explicitly revealed secondarily
- Operator surfaces (OPSURF-001): execution outcome, data completeness, governance result, and lifecycle/readiness are modeled as distinct status dimensions when all apply; they are not collapsed into one ambiguous status
- Operator surfaces (OPSURF-001): every mutating action communicates whether it changes TenantPilot only, the Microsoft tenant, or simulation only before execution
- Operator surfaces (OPSURF-001): dangerous actions follow configuration → safety checks/simulation → preview → hard confirmation where required → execute, unless a spec documents an explicit exemption and replacement safeguards
- Operator surfaces (OPSURF-001): workspace and tenant context remain explicit in navigation, actions, and page semantics; tenant surfaces do not silently expose workspace-wide actions
- Operator surfaces (OPSURF-001): each new or materially refactored operator-facing page defines a page contract covering persona, surface type, operator question, default-visible info, diagnostics-only info, status dimensions, mutation scope, primary actions, and dangerous actions
- Filament UI Action Surface Contract: for any new/modified Filament Resource/RelationManager/Page, define Header/Row/Bulk/Empty-State actions, ensure every List/Table has a surface-appropriate inspect affordance, remove redundant View when row click or identifier click already opens the same destination, keep standard CRUD/Registry rows to inspect plus at most one inline safe shortcut, group or relocate the rest to “More” or detail header, forbid empty bulk/overflow groups, require confirmations for destructive actions, write audit logs for mutations, enforce RBAC via central helpers (non-member 404, member missing capability 403), and ensure CI blocks merges if the contract is violated or not explicitly exempted
- Filament UI UX-001 (Layout & IA): Create/Edit uses Main/Aside (3-col grid, Main=columnSpan(2), Aside=columnSpan(1)); all fields inside Sections/Cards (no naked inputs); View uses Infolists (not disabled edit forms); status badges use BADGE-001; empty states have specific title + explanation + 1 CTA; max 1 primary + 1 secondary header action (see HDR-001); tables provide search/sort/filters for core dimensions; shared layout builders preferred for consistency
- Action-surface discipline (ACTSURF-001 / HDR-001): every changed
  surface declares one broad action-surface class; the spec names the
  one likely next operator action; navigation is separated from
  mutation; record/detail/edit pages keep at most one visible primary
  header action; monitoring/workbench surfaces separate scope/context,
  selection actions, navigation, and object actions; risky or rare
  actions are grouped and ordered by meaning/frequency/risk; any special
  type or workflow-hub exception is explicit and justified
- UI review workflow: native/custom classification, shared-family
  relevance, state-layer ownership, repository-signal treatment,
  exception path, and the active feature PR close-out entry stay
  explicit without duplicating the same decision across spec, plan,
  tasks, checklist, and close-out surfaces
- UI/Productization coverage (UI-COV-001): every spec completes the UI
  Surface Impact decision; changed reachable UI surfaces update
  `docs/ui-ux-enterprise-audit/` coverage artifacts with page archetype,
  design depth, repo-truth level, target pattern/mockup need,
  customer-safe review need, and dangerous-action review need; a checked
  `No UI surface impact` decision is used only with a short rationale

## Test Governance Check

> **Fill for any runtime-changing or test-affecting feature. Docs-only or template-only work may state concise `N/A` or `none`.**

- **Test purpose / classification by changed surface**: [Unit / Feature / Heavy-Governance / Browser / N/A]
- **Affected validation lanes**: [fast-feedback / confidence / heavy-governance / browser / profiling / junit / N/A]
- **Why this lane mix is the narrowest sufficient proof**: [Why the chosen classification and lanes fit the actual proving purpose]
- **Narrowest proving command(s)**: [Exact commands reviewers should run before merge]
- **Fixture / helper / factory / seed / context cost risks**: [none / describe]
- **Expensive defaults or shared helper growth introduced?**: [no / describe explicit opt-in path]
- **Heavy-family additions, promotions, or visibility changes**: [none / describe]
- **Surface-class relief / special coverage rule**: [standard-native relief / named special profile / N/A]
- **Closing validation and reviewer handoff**: [What must be re-run, what reviewers should verify, and what exact proof command they should rely on]
- **Budget / baseline / trend follow-up**: [none / describe]
- **Review-stop questions**: [lane fit / breadth / hidden cost / heavy-family risk / escalation]
- **Escalation path**: [none / document-in-feature / follow-up-spec / reject-or-split]
- **Active feature PR close-out entry**: [Guardrail / Exception / Smoke Coverage / N/A]
- **Why no dedicated follow-up spec is needed**: [Routine upkeep stays inside this feature unless recurring pain or structural lane changes justify a separate spec]

## Project Structure

### Documentation (this feature)

```text
specs/[###-feature]/
├── plan.md              # This file (/speckit.plan command output)
├── research.md          # Phase 0 output (/speckit.plan command)
├── data-model.md        # Phase 1 output (/speckit.plan command)
├── quickstart.md        # Phase 1 output (/speckit.plan command)
├── contracts/           # Phase 1 output (/speckit.plan command)
└── tasks.md             # Phase 2 output (/speckit.tasks command - NOT created by /speckit.plan)
```

### Source Code (repository root)
<!--
  ACTION REQUIRED: Replace the placeholder tree below with the concrete layout
  for this feature. Delete unused options and expand the chosen structure with
  real paths (e.g., apps/admin, packages/something). The delivered plan must
  not include Option labels.
-->

```text
# [REMOVE IF UNUSED] Option 1: Single project (DEFAULT)
src/
├── models/
├── services/
├── cli/
└── lib/

tests/
├── contract/
├── integration/
└── unit/

# [REMOVE IF UNUSED] Option 2: Web application (when "frontend" + "backend" detected)
backend/
├── src/
│   ├── models/
│   ├── services/
│   └── api/
└── tests/

frontend/
├── src/
│   ├── components/
│   ├── pages/
│   └── services/
└── tests/

# [REMOVE IF UNUSED] Option 3: Mobile + API (when "iOS/Android" detected)
api/
└── [same as backend above]

ios/ or android/
└── [platform-specific structure: feature modules, UI flows, platform tests]
```

**Structure Decision**: [Document the selected structure and reference the real
directories captured above]

## Complexity Tracking

> **Fill when Constitution Check has violations that must be justified OR when BLOAT-001 is triggered by new persistence, abstractions, states, or semantic frameworks.**

| Violation | Why Needed | Simpler Alternative Rejected Because |
|-----------|------------|-------------------------------------|
| [e.g., 4th project] | [current need] | [why 3 projects insufficient] |
| [e.g., Repository pattern] | [specific problem] | [why direct DB access insufficient] |

## Proportionality Review

> **Fill when the feature introduces a new enum/status family, DTO/presenter/envelope, persisted entity/table/artifact, interface/contract/registry/resolver, taxonomy/classification system, or cross-domain UI framework.**

- **Current operator problem**: [What present-day workflow or risk requires this?]
- **Existing structure is insufficient because**: [Why the current code cannot serve safely or clearly]
- **Narrowest correct implementation**: [Why this shape is the smallest viable one]
- **Ownership cost created**: [Maintenance, testing, cognitive load, migration, or review burden]
- **Alternative intentionally rejected**: [Simpler option and why it failed]
- **Release truth**: [Current-release truth or future-release preparation]